Keywords

Cancer screening aims to interfere with disease progression by detecting cancer at a point in its natural history when it is either curable or, if not curable, when treatment will extend life beyond what it would have been in the absence of cancer screening. The phrase screen detected and similar terms are used in this chapter but are a bit of a misnomer, as cancer screening tests are not the final arbiter of the presence of cancer. Screen detected is intended to mean that cancer screening initiated a process that led to the diagnosis of a cancer.

2.1 A Simple Model of the Natural History of Cancer

The natural history of cancer is complex and for the most part not well understood. Furthermore, there is great variability, even among tumors of the same organ. Figure 2.1 depicts how cancer progresses. It is an gross over-simplification of the process but it is a useful aid in explaining how cancer screening aims to interfere in the disease process.

Fig. 2.1
figure 1

Four phases of cancer progression

Full size image

Figure 2.1 displays four phases of cancer progression that are relevant to cancer screening. Cancer is present, asymptomatic, and not yet detectable by screening in Phase A. In Phase B (previously known as the detectable pre-clinical phase or DPCP), cancer is still asymptomatic but has characteristics that should, in the best of all possible worlds, make it detectable through cancer screening. Examples of characteristics are size and shedding of tumor cells that could be detected in a biospecimen. A cancer in Phase B may not be screen detected, however; the individual may not be screened or the test may give an inaccurate result due to its limitations. In Phase C, cancers come to clinical attention due to symptoms. Phase C includes cancers that are curable as well as those that are not. In Phase D, cancer causes death.

It is important to note that Phases A and B are a function of the cancer screening test. A cancer may be classified as being in Phase B if a technologically advanced test is used to screen, but in Phase A otherwise. For example, some lung cancers that can be detected with low dose computed tomography (LDCT) screening would not have been detectable with traditional two-dimensional chest x-ray screening given chest x-ray’s poorer resolution and capture of substantially less radiologic information. At a specific point in time, a cancer could be in Phase B if the cancer screening modality is LDCT and Phase A if the cancer screening modality is chest x-ray. Use of chest x-ray screening for lung cancer was common in the later decades of the 20th century but is no longer standard of care.

The purpose of the four phase model is not to demonstrate all possible paths that cancer or an individual with cancer can experience. It assumes that all cancers progress through each phase and do so only in a forward fashion, even though we know that some cancers regress or stall. It assumes that all cancers would be fatal if not treated, though experience tells us otherwise. It assumes that death due to causes other than the cancer of interest cannot occur. Even with those exclusions, the model is useful in conceptualizing our goals in cancer screening and provides a vocabulary that helps us discuss cancer screening.

Cancer screening attempts to shift the diagnosis of Phase C cancers to Phase B. Cancer screening is predicated on the belief that treatment at Phase B is more likely to lead to cure or extension of life than treatment at Phase C. If treatment at Phase B offers no prognostic advantage over treatment at Phase C, cancer screening will not lead to a reduction in cause-specific mortality. Treatment options may be more palatable, however, for Phase B cancers, and morbidity associated with the cancer and its treatment may be reduced. Then again, a diagnosis that occurs at an earlier time leads to more time spent as a cancer patient, which has psychological and clinical implications as discussed in the Benefits and Harms section of Chap. 1.

2.2 Three Important Phenomena in Screen Detection of Cancer

Lead time, length-biased sampling, and overdiagnosis are three terms that are used frequently in the assessment of cancer screening. They refer to the shift to an earlier date of diagnosis with cancer screening (lead time) and selection of a prognostically favorable (and thus non-random) sample of cancers (length-biased sampling and overdiagnosis). The remainder of this chapter explains these phenomena, while Chap. 5 describes how they complicate interpretation of cancer screening data.

The phrase length-biased sampling becomes awkward when we speak of bias due to length-biased sample. The phrase length time, which is sometimes used instead of length-biased sample, isn’t a better choice as it is not particularly descriptive. The phrase length-weighted sampling will be used instead in the remainder of this primer.

In the remainder of this primer, the phrase three cancer screening phenomena refer to lead time, length-weighted sampling, and overdiagnosis.

2.2.1 Lead Time

Screen-detected cancers are diagnosed at an earlier point in time than they would have been in the absence of cancer screening. Lead time is the amount of time by which the diagnosis date was advanced by cancer screening. It is that shifting of the diagnosis date to an earlier time that leads many to refer to cancer screening as early or earlier detection.

Lead time for an actual individual is impossible to calculate because we cannot know what the date of symptomatic diagnosis would have been in the absence of cancer screening. But for the purpose of illustration, we will pretend that we know that date. Lead time would be 3 months if, in the absence of cancer screening, a cancer would have been detected on June 1, 2018, but, in the presence of cancer screening, was diagnosed on March 1, 2018.

2.2.2 Length-Weighted Sampling

The term length-weighted sampling refers to the fact that the chance of screen detection is dependent on the length of time (sometimes referred to as sojourn time) the cancer remains in Phase B. The term sampling is used because cancer screening is merely a selection of some cancers (those that are screen detected) from the pool of all cancers. In elementary probability classes, sampling is often demonstrated using a jar of marbles. If the marbles are all of the same size, each marble has the same chance of selection. If the marbles are of different sizes, the chance that a given marble is selected is determined by its size, with chance of selection positively correlated with size of a marble. Cancer screening is similar to the latter situation; as the reader will see, one particular tumor characteristic, time spent in Phase B, drives the chance of detection.

Recall that not all cancers can be detected through screening. The chance that a cancer will be screen detected is dependent to differing degrees on many factors, including cancer characteristics, screenee characteristics, the cancer screening test, and screening interval. Screening interval refers to the amount of time between screens. An annual screening interval is used (or was used until we learned more about cancer progression) for a number of cancer screening tests.

Most notably, the chance of screen detection is inversely associated with the speed of tumor growth: faster growing cancers spend less time in Phase B and consequently have less time to be screen detected. A cancer that spends only 3 months in Phase B, for example, will have no opportunity to be detected on an annual screen, unless the annual screen happens to occur during that three-month window. A cancer that spends 2 years in Phase B will have two annual screens on which it can be detected. Cancers with a longer Phase B are assumed to be slower growing than those with a shorter Phase B, and cancers that are slower growing are assumed to have better prognosis. Therefore, length-weighted sampling leads to detection of cancers through screening that are expected to have better prognosis than those that are not detected though screening.

The term weighted is used to mean that the sample of cancers detected through screening will be skewed in favor of cancers with more favorable prognosis. In other words, the cancers that screening detects do not represent a random sample of all cancers.

Figure 2.2 depicts the fictional experience of three screenees, Mike, Molly, and Mary. The example demonstrates the interplay of screening interval and length of Phase B.

Fig. 2.2
figure 2

The interplay of Phase B length, a one-year screening interval, and cancer diagnosis. X indicates cancer diagnosis. Experience is fictional

Full size image

Mike, Molly, and Mary have Phase A cancers at the time of the first screen (month 0), so none of the three has cancer detected. Each cancer enters Phase B at month 6. Mike’s cancer enters Phase C at month 9, leading to a symptom-driven diagnosis prior to the second screen. Molly’s cancer is in Phase B at month 12, the time of her second screen, and the cancer is screen detected. Mary’s cancer also is in Phase B at month 12, but her cancer is missed at the second screen. Because Mary’s cancer is still in Phase B at the time of her third screen (24 months), there is another opportunity to detect it through screening, and that happens.

What would have happened had a different screening interval been used? A six-month interval could have led to screen detection for Mike, because the shorter the screening interval, the more likely that cancers with short Phase B lengths will be detected through cancer screening. The impact of a two-year interval on Molly’s experience depends on the length of Phase B: if it had been less than 18 months, Molly’s cancer could not have been screen detected.

2.2.3 Overdiagnosis

Overdiagnosis is the detection through cancer screening of cancers that never would have been diagnosed in the absence of cancer screening. These are cancers that, in the absence of screening, would not progress beyond Phase B during the lifetime of the screenee. The existence of overdiagnosis in cancer screening was once quite controversial, but nowadays many researchers and clinicians are open-minded to the possibility that some screen-detected cancers are diagnosed unnecessarily. Many also believe that screening for cancer of any site would result in overdiagnosis. The biggest controversy in overdiagnosis today surrounds its magnitude, which is further discussed in Chap. 9.

Overdiagnosis includes, but is not restricted to, the detection of indolent cancers. Indolent cancers are those that are screen detected (by definition, in Phase B), but in the absence of cancer screening and even in the longest of lifetimes, would have remained in Phase B, regressed to Phase A, or completely resolved. The screen detection of indolent cancers is an extreme example of length-weighted sampling. Because overdiagnosis refers to screen-detected cancers, cancer detected as a result of symptoms (Phase C cancers) cannot be overdiagnosed, even though they too, theoretically, can stall, regress, or resolve.

Non-indolent screen-detected cancers can be overdiagnosed if death, due to a cause other than the screen-detected cancer, occurs between the date of screen detection and the hypothetical date of symptom detection. The phrase competing cause of mortality is used to describe this scenario. In Fig. 2.3, John’s cancer is not overdiagnosed because he is alive on the day that the cancer would have been detected due to symptoms. James, however, dies soon after his cancer is diagnosed, and he is not alive on the day that it would have been detected due to symptoms had he lived. If not screen detected, James’ cancer would have been in Phase B when he died, and neither he nor his health care providers would have known of its existence.

Fig. 2.3
figure 3

Overdiagnosis due to death from other causes. Experience is fictional

Full size image

Today’s technology generally does not allow for the classification of a tumor as clearly indolent or not. In some instances, tumor characteristics can suggest a likely course, be it an innocuous or highly aggressive one. Death from a different cause soon after screen detection may suggest overdiagnosis associated with a competing cause of mortality, while death long after argues against it. Of course, uncertainty always exists because the life course an individual would have experienced in the absence of cancer screening is unknowable.